1. Field of the Invention
The present invention relates generally to telecommunications, and particularly to linking autonomous private network systems.
2. Technical Background
The Internet is comprised of many networks. An autonomous system, which may be part of a larger network, is a system that is administered by a single organization, regardless of whether it is public or private. Public network systems are identified by an IP (Internet Protocol) address. Public IP addresses are obtained from an Internet Registrar. An IP address is used to uniquely identify a host computer connection in an IP network. An IP address is a 32 bit binary number usually represented as 4 fields, each field representing an 8-bit number. They are often written as four decimal numbers separated by decimal points (e.g., for example, 125.115.21.6). However, since many private enterprise network applications do not require network layer connectivity outside the enterprise network, these private networks use unregistered IP addresses.
Private networks often use TCP/IP and its addressing capabilities for intra-enterprise communications. The Internet Assigned Numbers Authority (IANA) has reserved three blocks of IP address space for private internets in IETF RFC 1918. Private networks that do not require access to the Internet may use a block of IP addresses that are re-used by other private networks. IP addresses such as these are said to overlap. Examples of such private networks include private LANs, intranets, virtual private networks (VPNs), and etc.
Currently, it is becoming increasingly important to link private autonomous networks and provide application flows between them. Examples of such applications include business-to-business transactions, purchasing applications over VPNs, banking transactions, procurement applications, and etcetera. Another application relates to private network management. For example, a given company may implement a private intranet that may or may not have advanced features such as connections to the Internet, extranet capability, or whatever. Further, the company may not have the expertise to manage their private network, or the company may determine that the cost of managing their internal intranet is prohibitive. In order to solve the aforementioned problems, companies often outsource the management of their network to a network management company. From the management company's point of view, they want to manage as many customer networks as they can to derive a revenue stream that is as large as possible. Management charges are typically recurring, and are based on the size and number of routers in each customer's network. In order to accomplish this, the management company will attempt to link each of their Network Operations Centers (NOCs) to multiple customer networks.
FIG. 1 is an approach that has been considered. In this approach, the management company's NOC 2 is connected to a first customer network 3 and a second customer network 4 using redundant connections 12. NOC 2 also employs redundant network management systems (NMSs) 21, 22. This works perfectly well in the scenario depicted in FIG. 1 because the TCP/IP version 4-address space used by customer network 3 (10.1.1.0/24) does not overlap the IP address space used by customer network 4 (10.1.2.0/24). However, this approach has drawbacks.
A network management company desires to manage as many customer networks as they can in order to derive a larger revenue stream. The first problem relates to adding a large customer network that cannot be renumbered, and uses an address space that overlaps the address space of an existing customer network. Referring to FIG. 1, if the large network to be added (not shown) includes an address block of 10.1.1.0/24, it will conflict with the corresponding address block of customer network 3 because it is employing the same address space. One way of coping with this problem involves Network Address Translation (NAT).
NAT allows a router to act as an agent between a public network, such as the Internet, and a private autonomous network. This means that only a single, unique IP address is required to represent an entire group of host computers in the private network. Network Address Translation is used by a firewall, router or other computer disposed at the interface between the private network and the rest of the world. NAT is used to map an unregistered IP address to a registered IP address. This can be done on a one-to-one basis, or dynamically. However, NAT does not scale well. NAT requires an in-bound and outbound translation for every node in the coupled system. In a system with a large number of nodes, NAT becomes impractical. Thus, another solution is required.
The second problem with the network management scheme depicted in FIG. 1 relates to the issue of NOC geographic diversity. Assume that the outsourced NOC 2 for every domain is housed in the same physical facility (e.g., Chicago). If a fault condition occurs, such that the facility in Chicago becomes non-functional, all of the Network Management Systems (21, 22) will also be down for as long as the facility is non-functional. As long as the problem exists, the customer networks will not receive the services they contracted for. Obviously, this outcome is unacceptable to both parties.
What is needed is a way to link autonomous systems having overlapping IP addressing schemes without having to renumber, or use NAT. What is also needed is a way to couple geographically diverse and logically related autonomous systems. Furthermore, any solution to the above stated problems must be scalable, such that a large number of networks can be efficiently managed.